Recognizing material categories is one of the core challenges in robotic nuclear waste decommissioning. All nuclear waste should be sorted and segregated according to its materials, and then different disposal post-process can be applied. In this paper, we propose a novel transfer learning approach to learn boundary-aware material segmentation from a meta-dataset and weakly annotated data. The proposed method is data-efficient, leveraging a publically available dataset for general computer vision tasks and coarsely labeled material recognition data, with only a limited number of fine pixel-wise annotations required. Importantly, our approach is integrated with a Simultaneous Localization and Mapping (SLAM) system to fuse the per-frame understanding delicately into a 3D global semantic map to facilitate robot manipulation in self-occluded object heaps or robot navigation in disaster zones. We evaluate the proposed method on the Materials in Context dataset over 23 categories and that our integrated system delivers quasi-real-time 3D semantic mapping with high-resolution images. The trained model is also verified in an industrial environment as part of the EU RoMaNs project, and promising qualitative results are presented. A video demo and the newly generated data can be found at the project website (Supplementary Material).Many applications benefit from the use of multiple robots, but their scalability and applicability are fundamentally limited when relying on a central control station. Getting beyond the centralized approach can increase the complexity of the embedded software, the sensitivity to the network topology, and render the deployment on physical devices tedious and error-prone. https://www.selleckchem.com/products/epz-6438.html This work introduces a software-based solution to cope with these challenges on commercial hardware. We bring together our previous work on Buzz, the swarm-oriented programming language, and the many contributions of the Robotic Operating System (ROS) community into a reliable workflow, from rapid prototyping of decentralized behaviors up to robust field deployment. The Buzz programming language is a hardware independent, domain-specific (swarm-oriented), and composable language. From simulation to the field, a Buzz script can stay unmodified and almost seamlessly applicable to all units of a heterogeneous robotic team. We present the software structure of our solution, and the swarm-oriented paradigms it encompasses. While the design of a new behavior can be achieved on a lightweight simulator, we show how our security mechanisms enhance field deployment robustness. In addition, developers can update their scripts in the field using a safe software release mechanism. Integrating Buzz in ROS, adding safety mechanisms and granting field updates are core contributions essential to swarm robotics deployment from simulation to the field. We show the applicability of our work with the implementation of two practical decentralized scenarios a robust generic task allocation strategy and an optimized area coverage algorithm. Both behaviors are explained and tested with simulations, then experimented with heterogeneous ground-and-air robotic teams.How does AI need to evolve in order to better support more effective decision-making in managing the many complex problems we face at every scale, from global climate change, collapsing ecosystems, international conflicts and extremism, through to all the dimensions of public policy, economics, and governance that affect human well-being? Research in complex decision-making at an individual human level (understanding of what constitutes more, and less, effective decision-making behaviors, and in particular the many pathways to failures in dealing with complex problems), informs a discussion about the potential for AI to aid in mitigating those failures and enabling a more robust and adaptive (and therefore more effective) decision-making framework, calling for AI to move well-beyond the current envelope of competencies.Researchers investigating virtual/augmented reality have shown humans' marked adaptability, especially regarding our sense of body ownership; their cumulative findings have expanded the concept of what it means to have a body. Herein, we report the hand ownership illusion during "two views merged in." In our experiment, participants were presented two first-person perspective views of their arm overlapped, one was the live feed from a camera and the other was a playback video of the same situation, slightly shifted toward one side. The relative visibility of these two views and synchrony of tactile stimulation were manipulated. Participants' level of embodiment was evaluated using a questionnaire and proprioceptive drift. The results show that the likelihood of embodying the virtual hand is affected by the relative visibility of the two views and synchrony of the tactile events. We observed especially strong hand ownership of the virtual hand in the context of high virtual hand visibility with synchronous tactile stimulation.Quadruped robots require compliance to handle unexpected external forces, such as impulsive contact forces from rough terrain, or from physical human-robot interaction. This paper presents a locomotion controller using Cartesian impedance control to coordinate tracking performance and desired compliance, along with Quadratic Programming (QP) to satisfy friction cone constraints, unilateral constraints, and torque limits. First, we resort to projected inverse-dynamics to derive an analytical control law of Cartesian impedance control for constrained and underactuated systems (typically a quadruped robot). Second, we formulate a QP to compute the optimal torques that are as close as possible to the desired values resulting from Cartesian impedance control while satisfying all of the physical constraints. When the desired motion torques lead to violation of physical constraints, the QP will result in a trade-off solution that sacrifices motion performance to ensure physical constraints. The proposed algorithm gives us more insight into the system that benefits from an analytical derivation and more efficient computation compared to hierarchical QP (HQP) controllers that typically require a solution of three QPs or more. Experiments applied on the ANYmal robot with various challenging terrains show the efficiency and performance of our controller.